Withdrawal Time Estimate Accuracy Analysis

Analysis of the accuracy of withdrawal time estimates versus actual finalization times.

Key Statistics

Overview

Total Withdrawals: 279

Total Estimates: 19924

Error Metrics

Mean Error: -28.15 hours

Median Error: -23.74 hours

Mean Absolute Error: 28.40 hours

Positive error means actual time was later than estimated

Day-Based Accuracy

Correct Day: 57.8%

Within ±1 day: 81.2%

Within ±2 days: 95.7%

Within ±3 days: 100.0%

Within ±7 days: 100.0%

Estimate Direction

Early Estimates: 0.0% (actual later than estimated)

Late Estimates: 53.1% (actual earlier than estimated)

Excludes estimates within ±1 hour of actual time (considered correct)

Processing Patterns

Daily Processing

Mean Withdrawals per Day: 93.0

Median Withdrawals per Day: 79.0

Maximum Withdrawals per Day: 156

Days with Processing: 3

Withdrawals are processed in bulk once per day

Accuracy Over Time

Estimate Accuracy by Time Until Finalization

Mean Absolute Error (hours) by time remaining:

0-6 hours: 6.39 hours

6-12 hours: 8.60 hours

12-24 hours: 27.36 hours

24-48 hours: 57.93 hours

48-72 hours: 58.02 hours

72+ hours: nan hours

By Lead Time

Lead Time	Count	Mean Error (h)	Median Error (h)	Mean Abs Error (h)
0-24h	9171	0.26	0.26	0.26
24-48h	1808	-21.38	-23.74	21.43
48-72h	2600	-42.69	-47.74	42.69
72h+	6345	-65.19	-71.74	65.19

Visualizations

Error Distribution

Error Vs Leadtime

Notes:

Horizontal gaps represent regular update intervals
Points are jittered to better show density

Cumulative Accuracy

Accuracy Over Time

Daily Processing

Withdrawal Distribution

Individual Estimates

Notes

Analysis based on withdrawals that have both estimated finalization times and actual finalization times.
Positive errors mean that the actual finalization occurred later than estimated.
Negative errors mean that the actual finalization occurred earlier than estimated.
Data collected from DynamoDB table "lido_withdrawal_requests".
Withdrawals are processed in bulk once a day at roughly the same time.
Day-based accuracy metrics are more relevant than hour-based metrics due to the once-per-day processing pattern.
The processing patterns section focuses on unique withdrawals processed each day, not individual estimates.
For "Estimate Direction" statistics, estimates within ±1 hour of actual time are excluded (considered correct).
The "true lead time" reference line in the Error vs Leadtime chart shows the theoretical boundary with 24-hour steps.
Analysis timestamp: 2025-03-14 09:28:32 UTC